[jason810496]

Why

The Language SDK docs had no single view of what Go, Java and TypeScript each support, so users had to read three READMEs, and the per-SDK manifests drifted from the code.

What

• One generated matrix in the Language SDK docs, in three parts: features common to both authoring modes, Dag definition, and Python Dag with stub TaskHandler. A supported cell shows the SDK release that shipped it.
• The authoring modes are dag-definition and stub-task-handler. A capability that exists in both declares one entry per mode and appears in both parts.
• New Dag-definition dimensions for the primitives the SDKs ship: dag-args, explicit-dependencies, edge-labels, if-else-branching / switch-case-branching (replacing branching), task-loops, trigger-dag-run, dynamic-dag-generation, human-in-the-loop, asset-definition.
• asset-definition (declaring assets) is kept distinct from asset-scheduling (scheduling on asset events, including the AssetWatcher equivalent) and asset-state-store (the store keyed by an asset, not an authoring surface).
• New dimensions for gaps every SDK shares today: dag-callbacks (a Dag definition has to declare them itself; a Python Dag's callbacks already run), task-callbacks (declared per mode, since a stub task needs the Python-declared one to run) and listeners (the listener API OpenLineage consumes).
• A new update-lang-sdk-central-matrix prek hook regenerates the matrix from the manifests, like the per-SDK README hooks.

# <sdk>/capabilities.yaml
edge-labels:
  supported: true
  since: "3.4.0"
  sdk_version: "1.0.0"
mapped-tasks:
  dag-definition:
    supported: false
  stub-task-handler:
    supported: false

---

Was generative AI tooling used to co-author this PR?

• Yes, with help from Claude Code (Opus 4.8 and Opus 5) and Codex (GPT-5) following the guidelines

[jason810496]

All the comments are addressed in df39754.

I update the section with task-loops and all the features that I'm able to foresee, please let me know if there's other feature that I didn't mention.

[uranusjr]

PR description is out of date.

Can you set up your Claude skill or whatever to do this automatically going forward? This has happened more than a couple of times at this point and I don’t like checking this every time.

[jason810496]

All the comments are addressed in 4b7c04f, thanks for the review.

PR description is out of date.
Can you set up your Claude skill or whatever to do this automatically going forward? This has happened more than a couple of times at this point and I don’t like checking this every time.

Actually, I asked CC to do it explicitly right before I claim the PR is ready for review.

[jason810496]

Thanks for the review.